Robot caregiving should be personalized to meet the diverse needs of care recipients—assisting with tasks as needed, while taking user agency in action into account. In physical tasks such as handover, bathing, dressing, and rehabilitation, a key aspect of this diversity is the functional range of motion (fROM), which can vary significantly between individuals.
In this work, we learn to predict personalized fROM as a way to generalize robot decision-making in a wide range of caregiving tasks.
We propose a novel data-driven method for predicting personalized fROM using functional assessment scores from occupational therapy. We develop a neural model that learns to embed functional assessment scores into a latent representation of the user’s physical function. The model is trained using motion capture data collected from users with emulated mobility limitations. After training, the model is used to predict personalized fROM for new users without motion capture.
Through simulated experiments in environments inspired by handover, rehab, dressing, and bathing, we show that the personalized fROM predictions from our model enable the robot to provide personalized and effective assistance. We further evaluate our approach through a real-robot user study, showing that our method improves the user’s agency in action while providing the appropriate amount of assistance.
Coming Soon
GRACE, our novel data-driven method for predicting personalized fROM from functional assessment scores. First, we reformulate fROM prediction as a binary classification problem. Next, we learn a compact latent representation that captures a user’s functional assessment results. Finally, we train a neural network to predict the feasibility of a given joint configuration conditioned on the latent representation.
Left: Collecting data for a subject with emulated mobility limitations using ARAT. Center and Right: We collected functional assessments and fROM data (tracked by motion capture) with 11 subjects, each emulating four conditions.